AI-Driven Seabed Exploration: Expanding the Scope of Marine Citizen Science in SECOSTA

The Balearic Islands are home to a rich diversity of seabed, yet their distribution and evolution remain challenging to map and study, which has become a priority in the EU-marine strategy directive. The SECOSTA project contributes to bridge this gap by integrating advanced machine learning and low-cost technology with citizen participation, driving both scientific progress and community engagement. In this paper, we present an AI-Driven Seabed Exploration platform, an innovative tool built by secondary school students that integrates a submersible camera with GPS and a bathymetric probe to collect high-resolution seabed images alongside geo-referenced bathymetric data. The collected images are labelled by the students, so they can be used to train a convolutional neural network inspired by U-Net to identify seabed habitats. In terms of citizen science goals, the results obtained with this platform are promising, as the model can predict seabed habitat with skills comparable to expert-based identification. Also, the procedure ensures social engagement connecting with a population (secondary school students) which are often difficult to mobilize. The limitations and future improvements of the initiative are also discussed.